SPSS Directions - Graphical Assessment of Normality
These directions may seem super-simplistic to some of you. They are not meant to be insulting! They are just to take into account the many different levels of computer experience in this class.
Open SPSS. To do this, either double-click on the SPSS icon, or select Start, Programs, SPSS 8.0 for Windows (Student Version). Several options for entering data will appear. Choose the one that says "Type in data."
Enter data from the octopus mantle length data. Enter the actual mantle lengths and the normal scores and Z-scores. Name your columns. You can do this by double-clicking on the top of the column. SPSS has a bunch of rules about what you can name your columns. It is not case sensitive, so everything will appear in lower case, and it doesn't like symbols or spaces. Also, column names can be a maximum of 8 letters long. While you're naming your columns, you should make sure that they are the right data type - in this case they should be "numeric" - and you may want to increase the number of decimal places SPSS should keep track of, particularly for the Z-scores.
Make a histogram of the data. Go to Graphs, and choose Histogram. Several options will appear. Play with them until you get a histogram that makes some sense! To do that, choose the column of data you'd like to graph, specifically the mantle lengths (raw data). Note that there is an option to include a normal curve on your graph. Is this data normally distributed? left skewed? right skewed?
Create several scatter plots of the data. Go to Graphs, Scatterplot, Simple, and choose your x and y variables. Do this several ways using the mantle lengths, normal scores, and Z-scores. Note how the shapes of the graphs change depending on what values are on which axes. Also note the "stretch and squish" evident in the plot of the normal scores and Z values.
Make a Q-Q plot of the data. Go to Graphs, Q-Q plot, and choose the mantle length data. Compare this plot to the Q-Q plots in the notes. What do you notice? (Don't worry about the detrended normal scores.)
Transform your data. Go to Transform, Compute. Using the arrows, move the mantle data into the equation editor box. We will use the transformation y = log10(x +1). You will need to choose LOG10 from the list of operators in the box and use the up arrow to move it. When you are writing this equation, do NOT put in an equal sign - this is not Excel! Once you have calculated the transformed values, you can name your new variable. Then make another histogram and Q-Q plot. What do you notice?
χ2 Goodness of Fit Test
Let’s try to perform an actual test on some new data. We are interested in whether death is more likely in (or close to) one’s birth month than any other month of the year. There are 12 categories, and each represents the distance between the month of death and the month of birth. For example, “-6” means that death occurred 6 months prior to birth month, “0” means death in birth month and “6” means six months after birth month.
Read in the data set from the following web page by copying the link into Internet Explorer:
ftp://ftp.springer-ny.com/pub/supplements/voelkl/
In Internet Explorer a window should open with a bunch of icons. Find the “death.dat” file and right-click, then click: “Copy to Folder”
Save it to the desktop, then in SPSS go to FileàOpen. Select File Type “Tab-delimited (*.dat)”; find your file’s directory and select it, then click on “Open”. Do not read in variable name. You should find a single column of data that begins with “-6”. Each row in the data set represents the category that a single individual falls into, such that the total number of rows is equal to the total sample size. SPSS will count the numbers for each category and perform the Goodness of Fit test accordingly.
To do the test, Click on StatisticsàNonparametric testsàChi-Square. Click on the variable name (should be the only one listed) and the right arrow button to move it into the “Test Variable List” box. Click on OK. What results do you see?